I wrote:
> We have a couple votes for this patch and no one has spoken against it,
> so I'll go ahead and push it into HEAD.
BTW, I forgot to mention that while working on this patch I was thinking
it's past time to separate out the subquery support in allpaths.c into
its own file. With this patch, allpaths.c is 2363 lines, about 690 of
which are set_subquery_pathlist and subsidiary functions. While that
might not be quite tail-wagging-dog level, I think it's enough to justify
splitting that code out into a new file, say optimizer/path/subquerypath.c.
There are also about 630 lines devoted to appendrel path generation,
which might be thought enough to deserve a separate file of its own.
I'm a bit less excited about that though, mainly because the appendrel
logic has some not-quite-arms-length interactions with set_rel_size();
but there's certainly some case to be made for doing it.
Thoughts?
regards, tom lane